[nltk_data] Downloading package stopwords to /root/nltk_data... [nltk_data] Package stopwords is already up-to-date!
This document processes the outputs of the praise reward system and performs an analysis of the resulting token reward distribution.
Since praise gets valued on a scale, we can take a look at how often each value of the scale gets assigned by quantifiers. Note: This metric disregards scores of praise marked as a duplicate, since the score of the original is already being taken into account.
The ten highest rated contributions for this round were the following:
| Avg. score | To | Reason |
|---|---|---|
| 89.0 | kristofer#1475 | for their front-end and back-end development work on the praise dashboard |
| 84.67 | Tamarandom#9361 | Great 18 minute talk at ETH Denver (https://www.youtube.com/watch?v=7UvPCDbBKcQ). You must've prepped for hours for the talk and the slides you prepared. |
| 82.0 | Tamarandom#9361 | for supporting to accomodate people going to Amsterdam for Devconnect. Your clarity really helped us feel more calm regarding our staying plans. |
| 60.25 | GideonRo#3175 | the extraordinary visual explainers and workflows for the newcomer to the TEC and onboarding sensemaking from the many discussions. The skill you have to take multiple conversations and make them so concrete and clear is extraordinary. You really move the needle with your contributions to Communitas. Quanitfy this high, folks! Dozens of hours went into these three visual diagrams! |
| 59.33 | kristofer#1475 | for development on the praise dashboard and TEC praise bot |
| 56.67 | kristofer#1475 | for taking on the complex set of tickets to setup a docker deployment infrastructure for the praise app -- despite the technology being unfamilar to him |
| 55.47 | Tamarandom#9361 | all the crafty logistical work she did finding accommodation for 9 people in Amsterdam a week before the conference! |
| 50.67 | kristofer#1475 | for pushing out this new awesome bot! |
| 49.75 | iviangita#3204 | for being an incredible legal wg lead and singlehandedly pushing everything forward in this very critical working group |
| 49.75 | Vyvy-vi#5040 | for finish developing the /whoami and /help Praise commands! |
We can now take a look at the distribution of the received praise rewards. You can toggle the inclusion of the different sources by clicking on the legend.
We can also take a look at the amount of praise different users gave.
Now for something more fun: let's surface the top "praise flows" from the data. Thanks to @inventandchill for this awesome visualization! On one side we have the top 15 praise givers separately, on the other the top 25 receivers. The people outside the selection get aggregated into the "REST FROM" and "REST TO" categories.
Now let's take a closer look at the quantification process and the quantifiers:
To aid the revision process, we highlight disagreements between quantifiers.
This graphic visualizes controversial praise ratings by sorting them by the "spread" between the highest and lowest received score.
Please keep in mind that this is a visual aid. If there are several praise instances with similar spread and quant score, all but one end up "hidden" on the chart. For an exhaustive list, take a look at the exported file "praise_outliers.csv" .
Let's see how different quantifiers behaved by showing the range of praise scores they gave.
To interpret the box plot:
Bottom horizontal line of box plot is minimum value
First horizontal line of rectangle shape of box plot is First quartile or 25%
Second horizontal line of rectangle shape of box plot is Second quartile or 50% or median.
Third horizontal line of rectangle shape of box plot is third quartile or 75%
Top horizontal line of rectangle shape of box plot is maximum value.
Among 720 praises, 95 (12.94%) do not agree on duplication
Praise instances with disagreements in duplication are collected in 'results/duplication_examination.csv'. To compare, look at the last 4 columns: 'DUPLICATE MSG 1/2/3' and 'ORIGINAL MSG'.
Among 720 praises, 82 (11.17%) do not agree on dismissal
Praise instances with disagreements in dismissal are collected in'results/dismissal_disaggreed.csv'. You can further look into who dismissed and who did not.